import os
import cv2
import shutil

# ====================== 用户配置 ======================
SOURCE_DIR = "/media/hx/04e879fa-d697-4b02-ac7e-a4148876ebb0/dataset/1/zjdata15/2"      # 原图片文件夹
OUTPUT_DIR = "/media/hx/04e879fa-d697-4b02-ac7e-a4148876ebb0/dataset/1/zjdata15/2"    # 模糊图输出文件夹
BLUR_THRESHOLD = 180                     # 模糊阈值，越大越严格（100~300）
# ======================================================

IMG_EXT = [".jpg", ".jpeg", ".png", ".bmp", ".tiff", ".webp"]

os.makedirs(OUTPUT_DIR, exist_ok=True)

def is_blurry(image_path, threshold=150):
    """使用拉普拉斯方差检测是否模糊"""
    img = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE)

    if img is None:
        return False  # 读取失败跳过

    variance = cv2.Laplacian(img, cv2.CV_64F).var()

    return variance < threshold, variance


def scan_and_filter():
    total = 0
    bad = 0

    for filename in os.listdir(SOURCE_DIR):
        if not any(filename.lower().endswith(ext) for ext in IMG_EXT):
            continue

        img_path = os.path.join(SOURCE_DIR, filename)
        total += 1

        is_bad, score = is_blurry(img_path, BLUR_THRESHOLD)

        if is_bad:
            bad += 1
            shutil.copy(img_path, os.path.join(OUTPUT_DIR, filename))
            print(f"[低质量] {filename}  — 清晰度评分: {score:.2f}")
        else:
            print(f"[清晰]   {filename}  — 清晰度评分: {score:.2f}")

    print("\n============================")
    print(f"扫描总数: {total}")
    print(f"检测为模糊/马赛克: {bad}")
    print(f"已保存到: {OUTPUT_DIR}")
    print("============================\n")


if __name__ == "__main__":
    scan_and_filter()